Expectation Maximization with Bias-Corrected Calibration is Hard-To-Beat at Label Shift Adaptation

1 code implementation • ICML 2020 • Amr Mohamed Alexandari, Anshul Kundaje, Avanti Shrikumar

A limiting assumption of this algorithm is that p(y|x) is calibrated, which is not true of modern neural networks.

Medical Diagnosis

Fourier-transform-based attribution priors improve the interpretability and stability of deep learning models for genomics

no code implementations • NeurIPS 2020 • Alex Tseng, Avanti Shrikumar, Anshul Kundaje

To address these shortcomings, we propose a novel attribution prior, where the Fourier transform of input-level attribution scores are computed at training-time, and high-frequency components of the Fourier spectrum are penalized.

Adapting to Label Shift with Bias-Corrected Calibration

no code implementations • 25 Sep 2019 • Avanti Shrikumar, Amr M. Alexandari, Anshul Kundaje

Label shift refers to the phenomenon where the marginal probability p(y) of observing a particular class changes between the training and test distributions, while the conditional probability p(x|y) stays fixed.

Diabetic Retinopathy Detection Domain Adaptation +2

Maximum Likelihood with Bias-Corrected Calibration is Hard-To-Beat at Label Shift Adaptation

3 code implementations • 21 Jan 2019 • Amr Alexandari, Anshul Kundaje, Avanti Shrikumar

Label shift refers to the phenomenon where the prior class probability p(y) changes between the training and test distributions, while the conditional probability p(x|y) stays fixed.

Diabetic Retinopathy Detection Domain Adaptation +2

Technical Note on Transcription Factor Motif Discovery from Importance Scores (TF-MoDISco) version 0.5.6.5

1 code implementation • 31 Oct 2018 • Avanti Shrikumar, Katherine Tian, Žiga Avsec, Anna Shcherbina, Abhimanyu Banerjee, Mahfuza Sharmin, Surag Nair, Anshul Kundaje

TF-MoDISco (Transcription Factor Motif Discovery from Importance Scores) is an algorithm for identifying motifs from basepair-level importance scores computed on genomic sequence data.

Computationally Efficient Measures of Internal Neuron Importance

1 code implementation • 26 Jul 2018 • Avanti Shrikumar, Jocelin Su, Anshul Kundaje

We compare Neuron Integrated Gradients to DeepLIFT, a pre-existing computationally efficient approach that is applicable to calculating internal neuron importance.

A General Framework for Abstention Under Label Shift

1 code implementation • 20 Feb 2018 • Amr M. Alexandari, Anshul Kundaje, Avanti Shrikumar

In this work, we present a general framework for abstention that can be applied to optimize any metric of interest, that is adaptable to label shift at test time, and that works out-of-the-box with any classifier that can be calibrated.

Domain Adaptation General Classification +2

Learning Important Features Through Propagating Activation Differences

14 code implementations • ICML 2017 • Avanti Shrikumar, Peyton Greenside, Anshul Kundaje

Here we present DeepLIFT (Deep Learning Important FeaTures), a method for decomposing the output prediction of a neural network on a specific input by backpropagating the contributions of all neurons in the network to every feature of the input.

Interpretable Machine Learning

Not Just a Black Box: Learning Important Features Through Propagating Activation Differences

1 code implementation • 5 May 2016 • Avanti Shrikumar, Peyton Greenside, Anna Shcherbina, Anshul Kundaje

Note: This paper describes an older version of DeepLIFT.